xiaomi-sm8450/android_kernel_xiaomi_sm8450
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nouveau: simplify nouveau_dmem_migrate_vma

Browse Source 
Factor the main copy page to vram routine out into a helper that acts
on a single page and which doesn't require the nouveau_dmem_migrate
structure for argument passing.  As an added benefit the new version
only allocates the dma address array once and reuses it for each
subsequent chunk of work.

Link: https://lore.kernel.org/r/20190814075928.23766-8-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ralph Campbell <rcampbell@nvidia.com>
Tested-by: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
This commit is contained in:
Christoph Hellwig
2019-08-14 09:59:25 +02:00
committed by Jason Gunthorpe
parent bfe69ef94a
commit f268307ec7
1 changed files with 56 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files

186
drivers/gpu/drm/nouveau/nouveau_dmem.c
View File

@@ -44,8 +44,6 @@
#define DMEM_CHUNK_SIZE (2UL << 20) #define DMEM_CHUNK_SIZE (2UL << 20)
#define DMEM_CHUNK_NPAGES (DMEM_CHUNK_SIZE >> PAGE_SHIFT) #define DMEM_CHUNK_NPAGES (DMEM_CHUNK_SIZE >> PAGE_SHIFT)
struct nouveau_migrate;
enum nouveau_aper { enum nouveau_aper {
NOUVEAU_APER_VIRT, NOUVEAU_APER_VIRT,
NOUVEAU_APER_VRAM, NOUVEAU_APER_VRAM,
@@ -86,15 +84,6 @@ static inline struct nouveau_dmem *page_to_dmem(struct page *page)
return container_of(page->pgmap, struct nouveau_dmem, pagemap); return container_of(page->pgmap, struct nouveau_dmem, pagemap);
} }
struct nouveau_migrate {
struct vm_area_struct *vma;
struct nouveau_drm *drm;
struct nouveau_fence *fence;
unsigned long npages;
dma_addr_t *dma;
unsigned long dma_nr;
};
static unsigned long nouveau_dmem_page_addr(struct page *page) static unsigned long nouveau_dmem_page_addr(struct page *page)
{ {
struct nouveau_dmem_chunk *chunk = page->zone_device_data; struct nouveau_dmem_chunk *chunk = page->zone_device_data;
@@ -568,131 +557,66 @@ out_free:
drm->dmem = NULL; drm->dmem = NULL;
} }
static void static unsigned long nouveau_dmem_migrate_copy_one(struct nouveau_drm *drm,
nouveau_dmem_migrate_alloc_and_copy(struct vm_area_struct *vma, unsigned long src, dma_addr_t *dma_addr)
const unsigned long *src_pfns,
unsigned long *dst_pfns,
unsigned long start,
unsigned long end,
struct nouveau_migrate *migrate)
{ {
struct nouveau_drm *drm = migrate->drm;
struct device *dev = drm->dev->dev; struct device *dev = drm->dev->dev;
unsigned long addr, i, npages = 0; struct page *dpage, *spage;
nouveau_migrate_copy_t copy;
int ret;
/* First allocate new memory */ spage = migrate_pfn_to_page(src);
for (addr = start, i = 0; addr < end; addr += PAGE_SIZE, i++) { if (!spage || !(src & MIGRATE_PFN_MIGRATE))
struct page *dpage, *spage; goto out;
dst_pfns[i] = 0; dpage = nouveau_dmem_page_alloc_locked(drm);
spage = migrate_pfn_to_page(src_pfns[i]); if (!dpage)
if (!spage || !(src_pfns[i] & MIGRATE_PFN_MIGRATE)) return 0;
continue;
dpage = nouveau_dmem_page_alloc_locked(drm); *dma_addr = dma_map_page(dev, spage, 0, PAGE_SIZE, DMA_BIDIRECTIONAL);
if (!dpage) if (dma_mapping_error(dev, *dma_addr))
continue; goto out_free_page;
dst_pfns[i] = migrate_pfn(page_to_pfn(dpage)) | if (drm->dmem->migrate.copy_func(drm, 1, NOUVEAU_APER_VRAM,
MIGRATE_PFN_LOCKED | nouveau_dmem_page_addr(dpage), NOUVEAU_APER_HOST,
MIGRATE_PFN_DEVICE; *dma_addr))
npages++; goto out_dma_unmap;
}
if (!npages) return migrate_pfn(page_to_pfn(dpage)) |
return; MIGRATE_PFN_LOCKED | MIGRATE_PFN_DEVICE;
/* Allocate storage for DMA addresses, so we can unmap later. */ out_dma_unmap:
migrate->dma = kmalloc(sizeof(*migrate->dma) * npages, GFP_KERNEL); dma_unmap_page(dev, *dma_addr, PAGE_SIZE, DMA_BIDIRECTIONAL);
if (!migrate->dma) out_free_page:
goto error; nouveau_dmem_page_free_locked(drm, dpage);
migrate->dma_nr = 0; out:
return 0;
/* Copy things over */
copy = drm->dmem->migrate.copy_func;
for (addr = start, i = 0; addr < end; addr += PAGE_SIZE, i++) {
struct page *spage, *dpage;
dpage = migrate_pfn_to_page(dst_pfns[i]);
if (!dpage || dst_pfns[i] == MIGRATE_PFN_ERROR)
continue;
spage = migrate_pfn_to_page(src_pfns[i]);
if (!spage || !(src_pfns[i] & MIGRATE_PFN_MIGRATE)) {
nouveau_dmem_page_free_locked(drm, dpage);
dst_pfns[i] = 0;
continue;
}
migrate->dma[migrate->dma_nr] =
dma_map_page_attrs(dev, spage, 0, PAGE_SIZE,
PCI_DMA_BIDIRECTIONAL,
DMA_ATTR_SKIP_CPU_SYNC);
if (dma_mapping_error(dev, migrate->dma[migrate->dma_nr])) {
nouveau_dmem_page_free_locked(drm, dpage);
dst_pfns[i] = 0;
continue;
}
ret = copy(drm, 1, NOUVEAU_APER_VRAM,
nouveau_dmem_page_addr(dpage),
NOUVEAU_APER_HOST,
migrate->dma[migrate->dma_nr++]);
if (ret) {
nouveau_dmem_page_free_locked(drm, dpage);
dst_pfns[i] = 0;
continue;
}
}
nouveau_fence_new(drm->dmem->migrate.chan, false, &migrate->fence);
return;
error:
for (addr = start, i = 0; addr < end; addr += PAGE_SIZE, ++i) {
struct page *page;
if (!dst_pfns[i] || dst_pfns[i] == MIGRATE_PFN_ERROR)
continue;
page = migrate_pfn_to_page(dst_pfns[i]);
dst_pfns[i] = MIGRATE_PFN_ERROR;
if (page == NULL)
continue;
__free_page(page);
}
} }
static void static void nouveau_dmem_migrate_chunk(struct nouveau_drm *drm,
nouveau_dmem_migrate_finalize_and_map(struct nouveau_migrate *migrate) struct migrate_vma *args, dma_addr_t *dma_addrs)
{ {
struct nouveau_drm *drm = migrate->drm; struct nouveau_fence *fence;
unsigned long addr = args->start, nr_dma = 0, i;
nouveau_dmem_fence_done(&migrate->fence); for (i = 0; addr < args->end; i++) {
args->dst[i] = nouveau_dmem_migrate_copy_one(drm, args->src[i],
while (migrate->dma_nr--) { dma_addrs + nr_dma);
dma_unmap_page(drm->dev->dev, migrate->dma[migrate->dma_nr], if (args->dst[i])
PAGE_SIZE, PCI_DMA_BIDIRECTIONAL); nr_dma++;
addr += PAGE_SIZE;
} }
kfree(migrate->dma);
/* nouveau_fence_new(drm->dmem->migrate.chan, false, &fence);
* FIXME optimization: update GPU page table to point to newly
* migrated memory.
*/
}
static void nouveau_dmem_migrate_chunk(struct migrate_vma *args,
struct nouveau_migrate *migrate)
{
nouveau_dmem_migrate_alloc_and_copy(args->vma, args->src, args->dst,
args->start, args->end, migrate);
migrate_vma_pages(args); migrate_vma_pages(args);
nouveau_dmem_migrate_finalize_and_map(migrate); nouveau_dmem_fence_done(&fence);
while (nr_dma--) {
dma_unmap_page(drm->dev->dev, dma_addrs[nr_dma], PAGE_SIZE,
DMA_BIDIRECTIONAL);
}
/*
* FIXME optimization: update GPU page table to point to newly migrated
* memory.
*/
migrate_vma_finalize(args); migrate_vma_finalize(args);
} }
@@ -704,38 +628,40 @@ nouveau_dmem_migrate_vma(struct nouveau_drm *drm,
{ {
unsigned long npages = (end - start) >> PAGE_SHIFT; unsigned long npages = (end - start) >> PAGE_SHIFT;
unsigned long max = min(SG_MAX_SINGLE_ALLOC, npages); unsigned long max = min(SG_MAX_SINGLE_ALLOC, npages);
dma_addr_t *dma_addrs;
struct migrate_vma args = { struct migrate_vma args = {
.vma = vma, .vma = vma,
.start = start, .start = start,
}; };
struct nouveau_migrate migrate = {
.drm = drm,
.vma = vma,
.npages = npages,
};
unsigned long c, i; unsigned long c, i;
int ret = -ENOMEM; int ret = -ENOMEM;
args.src = kzalloc(sizeof(long) * max, GFP_KERNEL); args.src = kcalloc(max, sizeof(args.src), GFP_KERNEL);
if (!args.src) if (!args.src)
goto out; goto out;
args.dst = kzalloc(sizeof(long) * max, GFP_KERNEL); args.dst = kcalloc(max, sizeof(args.dst), GFP_KERNEL);
if (!args.dst) if (!args.dst)
goto out_free_src; goto out_free_src;
dma_addrs = kmalloc_array(max, sizeof(*dma_addrs), GFP_KERNEL);
if (!dma_addrs)
goto out_free_dst;
for (i = 0; i < npages; i += c) { for (i = 0; i < npages; i += c) {
c = min(SG_MAX_SINGLE_ALLOC, npages); c = min(SG_MAX_SINGLE_ALLOC, npages);
args.end = start + (c << PAGE_SHIFT); args.end = start + (c << PAGE_SHIFT);
ret = migrate_vma_setup(&args); ret = migrate_vma_setup(&args);
if (ret) if (ret)
goto out_free_dst; goto out_free_dma;
if (args.cpages) if (args.cpages)
nouveau_dmem_migrate_chunk(&args, &migrate); nouveau_dmem_migrate_chunk(drm, &args, dma_addrs);
args.start = args.end; args.start = args.end;
} }
ret = 0; ret = 0;
out_free_dma:
kfree(dma_addrs);
out_free_dst: out_free_dst:
kfree(args.dst); kfree(args.dst);
out_free_src: out_free_src: